HW4 Problem 2 plotly

load dataset and start making plots

data(rest_inspec)

ri <-
  rest_inspec |>
  mutate(
    score = suppressWarnings(as.numeric(score)),
    long  = as.numeric(longitude),
    lat   = as.numeric(latitude)
  ) |>
  # ✅ rename cuisine_description to cuisine here
  rename(cuisine = cuisine_description) |>
  select(
    camis, dba, boro, cuisine, grade, score, lat, long, inspection_date
  ) |>
  filter(
    !is.na(score),
    !is.na(long), !is.na(lat),
    long < -70, long > -75,   # keep within NYC bounds
    lat  > 40, lat  < 41,
    boro == "Manhattan",
    score >= 0, score <= 40
  )

# ✅ now this will work
top_cuisines <- ri |>
  count(cuisine, sort = TRUE) |>
  slice_head(n = 12) |>
  pull(cuisine)

ri_small <- ri |> filter(cuisine %in% top_cuisines)

Plotly scatterplot: Scatterplot of restaurant locations and inspection scores

ri_small |>
mutate(text_label = str_c(
dba, " — ", cuisine,
"\nScore: ", score,
ifelse(!is.na(grade), str_c(" (Grade ", grade, ")"), ""),
"\nBorough: ", boro
)) |>
plot_ly(
x = ~long, y = ~lat, type = "scatter", mode = "markers",
color = ~score, colors = "viridis",
text = ~text_label, alpha = 0.6
) |>
layout(xaxis = list(title = "longitude"),
yaxis = list(title = "latitude"))

This scatterplot displays the spatial distribution of restaurant inspections in Manhattan. Each point is an individual inspection, with color showing the inspection score (lower = better). The points trace the Manhattan street grid, confirming correct filtering. Colors appear evenly mixed across the map — suggesting no clear geographic clustering of better or worse scores. Cleanliness levels are fairly consistent across the borough.

Plotly boxplot: Boxplot of inspection scores by cuisine

ri_small |>
mutate(cuisine = fct_reorder(cuisine, score, .fun = median, na.rm = TRUE)) |>
plot_ly(y = ~score, color = ~cuisine, type = "box", colors = "viridis") |>
layout(yaxis = list(title = "Inspection score (lower = better)"))

This boxplot compares inspection score distributions across the top 12 cuisines in Manhattan. Lower median scores represent better sanitary conditions. Cuisines like Café/Coffee/Tea and American have relatively low medians and compact ranges — generally strong compliance. Meanwhile, Latin and Delicatessen cuisines show higher medians and wider spreads, implying more inconsistent cleanliness. Overall, different cuisines exhibit moderate variability, likely due to differences in restaurant size and preparation complexity.

Plotly barchart: Number of inspections by cuisine in our subset

ri_small |>
count(cuisine) |>
mutate(cuisine = fct_reorder(cuisine, n)) |>
plot_ly(x = ~cuisine, y = ~n, color = ~cuisine, type = "bar", colors = "viridis") |>
layout(xaxis = list(title = "Cuisine"),
yaxis = list(title = "Number of inspections"))

This bar chart shows the frequency of inspections by cuisine type. Delicatessen, Café/Coffee/Tea, and Latin cuisines dominate the sample, reflecting their popularity in Manhattan. Less frequent categories like French and Bakery appear much smaller. This context is important when interpreting the boxplot: cuisines with few inspections may display greater variability simply due to limited data.